Course of Depressive Symptoms After Myocardial Infarction and Cardiac
Prognosis: A Latent Class Analysis
KIRSTEN I. KAPTEIN, MD, PETER DE JONGE, PHD, ROB H. S. VAN DEN BRINK, PHD, AND JAKOB KORF, PHD
Objective: The presence of depressive symptoms after myocardial infarction (MI) is a risk factor for new cardiovascular events.
The importance of the course of post-MI depressive symptoms for cardiac prognosis is not clear. We therefore set out to investigate
whether different courses of post-MI depressive symptoms can be identified and determine their associations with cardiac events.
Methods: Data were derived from the Depression after Myocardial Infarction (DepreMI) study, a naturalistic follow-up study of
patients admitted for an MI in four hospitals in the Netherlands (N ? 475). Scores on the Beck Depression Inventory (BDI) during
hospitalization and at 3, 6, and 12 months post-MI were analyzed. Using latent class analysis (LCA), we identified classes
characterized by distinctive courses of depressive symptoms and then examined their link to cardiac prognosis. Results: The
prevalence of significant depressive symptoms ranged from 22.7% to 25.5% throughout the post-MI year. Five distinct courses were
found: no depressive symptoms (56.4%), mild depressive symptoms (25.7%), moderate and increasing depressive symptoms
(9.3%), significant but decreasing depressive symptoms (4.6%), and significant and increasing depressive symptoms (4.0%).
Subjects in this last class had, statistically, a significantly higher risk for a new cardiovascular event compared with subjects without
depressive symptoms (hazard ratio (HR) ? 2.73; p ? .01). Controlling for baseline cardiac status and sociodemographic data did
not alter the association (HR ? 2.46; p ? .03). Conclusions: Post-MI depressed subjects with significant and increasing depressive
symptoms are at particular risk of new cardiac events. This subgroup may be most suited for evaluation of the effects of
antidepressant treatment on cardiac prognosis. Key words: depression, myocardial infarction.
BDI ? Beck Depression Inventory; BIC ? Bayesian Information
Criteria; CABG ? coronary artery bypass graft; CIDI ? Composite
International Diagnostic Interview; DepreMI ? Depression after
Myocardial Infarction; HR ? hazard ratio; ICD-10 ? International
Classification of Diseases, Version 10; LCA ? latent class analysis;
LVEF ? left ventricular ejection fraction; MI ? myocardial infarc-
tion; PTCA ? percutaneous transluminal coronary angioplasty.
Several studies have found post-MI depression to be a risk
factor for reinfarction and death (5–9). It remains unclear
whether the course of depressive symptoms is important to
cardiovascular prognosis. Such information could help in the
interpretation of the nature of the effect of depression on
cardiac prognosis and might also guide treatment efforts for
post-MI depression. In the ENRICHD study (10), for example,
only limited effects of active treatment of post-MI depression
on depressive symptoms and no effects on cardiovascular
prognosis compared with usual care were found. In part, this
seems to be due to a spontaneous recovery of a considerable
proportion of post-MI depressed patients in the usual care
group. A detailed analysis of the course of post-MI depressive
symptoms might reveal subgroups with varying courses of
depressive symptoms and different risks for new cardiac
Although in general psychiatry the importance of the
course of depression has been acknowledged (11–13), less
attention has been paid to the course of depressive symptoms
after MI. Only a few studies have investigated the develop-
pproximately 25% of myocardial infarction (MI) patients
develop depressive symptoms in the year after MI (1–5).
ment of post-MI depressive symptoms after discharge; in
these studies, it was found that on average depressive symp-
toms persisted or decreased somewhat over the following
months post-MI (4,14,15). To our knowledge, however, no
attempts have been made to form empirically derived classes
of subjects based on their post-MI course of depressive symp-
toms. In this study, we will therefore apply latent class anal-
ysis (LCA) (16–18) for this purpose. LCA has been used to
describe depressive symptoms, major depression, and psy-
chotic and melancholic depressions (19–22) in cross-sectional
studies. However, with LCA, longitudinal data can also be
analyzed (23–25). In the present study, we set out to classify
MI patients according to their course of depressive symptoms
and evaluate their cardiac prognosis.
The methods of the Depression after Myocardial Infarction (DepreMI)
study have been described in detail elsewhere (26,27) and are briefly de-
Design and Patients
The DepreMI study is a naturalistic follow-up study of the impact of
depressive symptoms on cardiac prognosis in MI patients in four hospitals in
the North of The Netherlands. Patients admitted for an MI during September
1997 to September 2000 were included and followed until April 2002.
Patients received usual aftercare for their MI and depressive symptoms.
Inclusion criteria were a) chest pain for at least 20 minutes, b) creatinine
phosphokinase levels 100% above normal or creatinine phosphokinase MB
levels above 10%, and c) presence of new pathological Q waves on the
electrocardiogram in at least two leads. Exclusion criteria were life expect-
ancy of less than a year (because of noncardiac condition), too poor physical
condition according to hospital staff, cognitive dysfunction, inability to speak
or read Dutch, occurrence of an MI in patients admitted for another reason,
and follow-up visits scheduled in a nonparticipating hospital. All participating
patients signed an informed consent form. The study protocol was approved
by the ethics committee review board at the participating hospitals.
The Beck Depression Inventory (BDI) (28) was scored during hospital
stay and at 3, 6, and 12 months post-MI. The BDI is a widely used 21-item
self-report measure to assess the presence and severity of depressive symp-
toms. Each symptom is rated from 0 to 3, representing increasing levels of
From the Department of Psychiatry (K.I.K., P.d.J., R.H.S.v.d.B., J.K.) and
Department of Internal Medicine (P.d.J.), University Medical Center Gro-
ningen, University of Groningen, The Netherlands.
Address correspondence and reprint requests to Peter de Jonge, PhD,
Oostersingel 59, P.O. Box 30001, Groningen, 9700 RB Netherlands. E-mail:
Received for publication October 20, 2005; revision received March 7,
Copyright © 2006 by the American Psychosomatic Society
Psychosomatic Medicine 68:662–668 (2006)
severity, with 0 representing absence of the symptom. A score of 10 or more
is generally accepted as having significant depressive symptoms, and a score
of 20 or more is regarded as having depressive symptoms of at least moderate
intensity. At approximately 3 and 12 months post-MI, patients underwent the
Composite International Diagnostic Interview (CIDI) to assess the presence of
a depressive disorder according to International Classification of Diseases,
Version 10 (ICD-10) diagnostic criteria.
Sociodemographic data (gender, age at MI, living alone or not, smoker or
nonsmoker, and body mass index) and information on history of depression
(occurrence either more than 1 year before the MI or less than 1 year before
MI) were collected during hospitalization. At approximately 3 months post-
MI, neuroticism and extraversion were measured with the Eysenck Person-
ality Questionnaire (29). Vital exhaustion was also measured at approximately
3 months post-MI, using the Maastricht Questionnaire (MQ) (30).
Severity of the index MI was represented by the left ventricular ejection
fraction (LVEF), maximum cpk-mb level, revascularization, arrythmia, and
presence of heart failure. LVEF was assessed during hospitalization by
echocardiography, magnetic resonance imaging, angiography, or nuclear ra-
diography. For reasons of comparability, LVEF was dichotomized as ?40%
or ?40%. Revascularization (percutaneous transluminal coronary angioplasty
(PTCA) or coronary artery bypass graft (CABG)) and the occurrence of an
arrhythmic event were all recorded during hospitalization. Killip class (di-
chotomized as 0–1 versus ?2) was used to assess the presence of heart failure
Study end points included cardiovascular mortality and cardiac-related
readmissions after discharge from the hospital. Information on potential end
points was collected from hospital records and the patient’s primary care
physician. Two cardiologists independently evaluated the nature (cardiovas-
cular or not) and onset of the end points. Decisions were required to be
unanimous. Follow-up time was defined as the time from index MI until a) the
occurrence of cardiovascular complication, b) death of the patient for non-
cardiovascular reasons, c) refusal of the patient to participate any further, and
d) end of follow-up time. Mean follow-up duration was 2.5 years.
When one of the four BDI assessments was missing, we used the single
multiple-imputation method (31–33) to replace the missing BDI. Single
imputation can be used to replace missing data when less than 10% of the data
are missing, without biasing the results (34).
Latent Class Analysis
We applied LCA (16–18) to the 4 BDI assessments of the entire cohort.
LCA is a statistical model-fitting method to identify different classes of
subjects within a given data set. LCA assumes unobserved latent variables to
explain the associations among observed scores and can be seen as a cate-
gorical equivalent of factor analysis, which assumes continuously distributed
latent variables. Instead of giving a particular true solution, LCA produces
several solutions with relative fit indices. LCA computes two sets of param-
eters. The first set is the latent class probabilities or class prevalences. The
other set of parameters is called the conditional probabilities and estimates the
probability of the observed variables, given that the individual is a member of
that class. The conditional probabilities are analogous to the factor loadings in
factor analysis. The Bayesian information criteria (BIC ? log (L) ? 0.05 ?
log (n) ? k, where k is the number of parameters) (35) are generally used for
the goodness of fit to determine the optimal number of groups. The smallest
BIC value gives the best fit, but a difference of less than 6 will favor the
higher BIC value. The null model is a model for one single class, i.e., the
whole cohort belonging to the same latent class. This model is rejected when
models with two or more parameters result in better fit indices. LCA requires
no intercorrelations between the assessments within classes, because it is
assumed that all correlations between the assessments can be explained by the
latent classes. We therefore assessed the correlations between the four BDI
assessments of the entire cohort and within each of the latent classes.
Comparison Between Classes
The resulting latent classes were compared on the following variables:
sociodemographic data (gender, age at MI, living alone or not, smoker or
nonsmoker, and body mass index), history of depression (depression more
than 1 year before MI and depression less than 1 year before MI), baseline
cardiologic data (history of MI, LVEF ?40%, Killip class ?2, max cpk-mb,
arrhythmic event during admission, PTCA and CABG during admission),
psychometric data (neuroticism, extraversion, and vital exhaustion), and
post-MI depression characteristics based on BDI and CIDI. For these analy-
ses, we used the Pearson chi square (categorical variables) and ANOVA
(continuous variables). Statistical significance was evaluated using two-sided,
p ? .05 levels.
For each latent class, we calculated a Kaplan-Meier survival curve (36) to
compare the rate of cardiac events (fatal or nonfatal) among the classes.
Differences between the curves were tested with the log rank test. After that,
every single class was compared against the class of nondepressed subjects by
Cox regression analysis. We tested confounding due to the severity of the MI,
gender, living alone, and history of MI by adding these variables in the Cox
models. We included these potential confounders because of their relation
with cardiovascular prognosis and with depression. We did not use the risk
factors smoking and body mass index (BMI), due to a considerable number of
missing values (44 and 86 cases).
Inclusion criteria for the DepreMI study were met by 1166
patients, but 284 (24%) were then ineligible based on the
exclusion criteria. Of the remaining 882 patients, 528 (60%)
gave informed consent. Four hundred thirty-eight patients
completed all BDI assessments, and 37 patients missed one
BDI assessment, which is acceptable for conducting single
imputation (33). Figure 1 shows the flow chart.
The 475 patients in our analyses consisted of 90 women
and 385 men, with a mean age of 60.6 years. Based on the
criterion of BDI ?10, the prevalence of significant depressive
symptoms throughout the post-MI year was 22.7% during
hospitalization, 23.8% 3 months post-MI, 25.5% 6 months
post-MI, and 24.8% 12 months post-MI. Based on the crite-
rion of BDI ?20, the percentages were 2.9%, 4.8%, 4.9%, and
5.5%, respectively. Based on the CIDI interviews adminis-
tered at 3 and 12 months post-MI, 116 of 461 patients (25.2%)
met the ICD-10 criteria for depressive disorder during the
post-MI year. One hundred twelve (23.6%) patients experi-
enced a cardiovascular event during follow-up, of which 21
(4.4%) were fatal.
For the 1-, 2-, 3-, 4-, and 5-class solutions, the BIC values
were decreasing for every additional class while still resulting
in classes with sufficient numbers of subjects (N ? 20). The
6-class solution gave an even smaller BIC value but contained
two small classes of 6 and 7 subjects, respectively. The BIC
value for the 7-class solution could not be calculated, due to a
class of 0 subjects. The 8-class solution gave a larger BIC
value and was therefore rejected. We chose the 5-class solu-
tion as best because it had the smallest BIC value compared
with solutions with 1 to 4 classes while still having classes
with sufficient subjects for further analyses (smallest class:
COURSE OF DEPRESSIVE SYMPTOMS
663Psychosomatic Medicine 68:662–668 (2006)
N ? 19). Table 1 shows the BIC values of the LCA classes.
The classes were characterized as 1) no depressive symptoms
(N ? 268, 56.4%), 2) mild depressive symptoms (N ? 122,
25.7%), 3) moderate, increasing depressive symptoms (N ?
44, 9.3%), 4) significant but decreasing depressive symptoms
(N ? 22, 4.6%), and 5) significant and increasing depressive
symptoms (N ? 19, 4.0%). The course of depressive symp-
toms of the five latent classes is shown in Figure 2. Testing the
correlations among BDI scores for the total cohort and for
the 5-class solution resulted in significant correlations among
the BDI scores for the total cohort (varying from 0.6 to 0.8)
but not within the classes (varying from ?0.2 to 0.4), indi-
cating that this model fits the data well.
Comparison Between Classes
As shown in Table 2, a comparison of classes on several
characteristics revealed that a majority of subjects in classes 3
(moderate, increasing), 4 (significant, decreasing), and 5 (sig-
nificant, increasing) developed a depressive disorder during
the post-MI year. In addition, class 5 (significant, increasing) is
characterized by severe depressive symptoms at all follow-up
depressive symptomatology during hospitalization, but the pres-
ence of at least moderate symptoms of depression (indicated
by BDI ?20) disappears after 3 months post-MI.
The Kaplan-Meier survival curves to describe the time until
a cardiovascular event for the five latent classes resulted in a
significant overall difference in survival among the classes
(log rank: 10.79, df ? 4, p ? .029) (Figure 3). When com-
paring classes 2 to 5 against class 1 (no depressive symptoms),
all classes tended toward higher rates of new cardiac events
(hazard ratio (HR) class 2: 1.89, class 3: 1.86, class 4: 1.43,
class 5: 2.73). However, only the class 5 of subjects with
significant and increasing depressive symptoms had a signif-
icantly higher rate (HR: 2.73, p ? .01) (Table 3). Controlling
for baseline MI severity and additional risk factors did not
alter the association.
The prevalence of depressive symptoms in our sample of
MI patients was common: 22.7% developed significant de-
pressive symptoms during hospitalization, which is in line
with previous reports (2,3,5,7,8). In contrast to most of the
existing literature, we were also able to study the course of
depression throughout the post-MI in more detail. We found
that the presence of significant depressive symptoms was
relatively stable, ranging from 23.8% at 3 months to 24.8% at
12 months post-MI. About a quarter of the subjects (25.2%)
developed a depressive disorder fulfilling ICD-10 criteria dur-
ing the post-MI year.
By analyzing the course of depressive symptoms in more
detail, we were able to identify five distinct classes: Class 1
included the 56.4% of subjects without depressive symptoms;
class 2, 25.7% of subjects who developed mild depressive
Figure 1. Flowchart. DepreMI study. *Beck Depression Inventory (BDI) assessments during hospitalization and at 3, 6, and 12 months post-MI.
TABLE 1. Bayesian Information Criteria (BIC) Values for the
Latent Class Analysis Classes
Class Solution BIC Value
aCould not be calculated due to a 0-subject class.
K. I. KAPTEIN et al.
664Psychosomatic Medicine 68:662–668 (2006)
symptoms; class 3, 9.3% with moderate but increasing depres-
sive symptoms; class 4, 4.6% with significant but decreasing
depressive symptoms; and class 5, 4.0% of the subjects with
significant and increasing depressive symptoms. Of the sub-
jects in class 5, a majority developed a post-MI depressive
disorder during the post-MI year (63.2%), which was rather
similar for classes 3 (69.2%) and 4 (54.4%). Remarkable
about subjects in class 5, however, was the persistence of
symptoms which were of at least moderate intensity (BDI
?20) during all follow-up assessments. Subjects in this class
also had the highest rate of new cardiovascular events com-
pared with subjects with no depressive symptoms (HR ?
2.73). This difference was not explained by the severity of the
MI or additional risk factors (HRadjusted? 2.46).
Some disagreement still exists about whether the presence
of post-MI depressive symptoms is a causal risk factor for
cardiac prognosis (4,37,38), although in a recent meta-analysis
(9), consistent associations were reported. Our current find-
ings underscore the need to look more closely to identify
subtypes of post-MI depression based on the course of symp-
toms. Not much is known yet about differential courses of
post-MI depression. On average, depressive symptoms seem
to be stable during the post-MI year, which is reported else-
where (e.g., 4,11,39) and is replicated in our present study.
However, when looking at the specific courses of depressive
symptoms, a significantly increased rate for new cardiac
events was found only for one class of subjects with signifi-
cant depressive symptoms that increased during the post-MI
year. In contrast, another class of subjects with significant
depressive symptoms during hospitalization whose symptoms
decreased during the post-MI year did not have an elevated
risk. A third class also seems to be of interest: subjects with
elevated BDI scores during hospitalization that slowly seem to
increase during the post-MI year. In this class of subjects, a
majority (69.8%) experienced a depressive disorder fulfilling
diagnostic criteria. Moreover, an increased rate of new cardiac
events (HR ? 1.86) was found, with a borderline significance.
Many of these patients will be missed when depression is only
screened for during hospitalization.
The following limitations and strengths of our study need
to be mentioned. Our choice for the 5-class solution, although
the 6-class solution had a lower and therefore better BIC
value, is debatable. The 5-class solution, however, gave suf-
ficient subjects in each of the classes instead of two small
subgroups of 7 and 6 subjects found in the 6-class solution. A
strength of our study is the use of a large number of subjects
(475) who experienced an MI. Another strength is the use of
repeated BDI and CIDI assessments and the LCA method to
identify different courses of depressive symptoms after MI,
which has not been applied previously in this context. Among
the study limitations, the considerable proportion of excluded
patients during the inclusion phases of the study should be
considered. Specifically, the number of patients who did not
give informed consent may have resulted in an underrepre-
sentation of patients with post-MI depression. In order to see
if our study was still representative of the population, we
therefore compared our data with results reported in previous
studies. Generally, an estimated prevalence of 15% to 25% of
post-MI depression is reported, so our finding of 25.2% seems
in line. We therefore do not expect that major bias has been
introduced. Finally, as this is a new approach to describe
courses of post-MI depressive symptoms, our results need to
be confirmed by others using this methodology.
Insight in the course of depressive symptoms after MI may
be important for clinical practice. Our findings suggest that MI
patients need to be followed up for depressive symptoms also
after hospitalization for their index MI because in some pa-
tients, depressive symptoms may increase only then. It re-
mains unclear whether this represents a deteriorating medical
status or a negative development of depressive symptoms that
stands by itself. On the other hand, some of the subjects who
experience depressive symptoms during hospitalization may
Figure 2. Mean BDI scores of the five latent classes during the post-MI year.
COURSE OF DEPRESSIVE SYMPTOMS
665 Psychosomatic Medicine 68:662–668 (2006)
Characteristics of the Latent Classes Based on the Course of Post-MI Depressive Symptoms
Male sex (%)
Age at MI timea
Living alone (%)
History of depression
?1 yr (%)
?1 yr (%)
Previous MI (%)
LVEF ?40% (%)
Killip class ?2 (%)
Arrhythmic event (%)
ICD-10 depression (%)
BDI ?10, during hosp. (%)
BDI ?10, 3 months post-MI (%)
BDI ?10, 6 months post-MI (%)
BDI ?10, 12 months post-MI (%)
BDI ?20, during hosp. (%)
BDI ?20, 3 months post-MI (%)
BDI ?20, 6 months post-MI (%)
BDI ?20, 12 months post-MI (%)
Symptoms discussed with any healthcare
Symptoms treated by any healthcare worker (%)
Symptoms treated by mental health care
Antidepressant medication prescribed (%)
MI ? myocardial infarction; BMI ? body mass index (weight in kg/height in m2); LVEF ? left ventricular ejection fraction; PTCA ? percutaneous transluminal coronary angioplasty; CABG ? coronary
artery bypass graft; BDI ? Beck Depression Inventory.
aANOVA for comparison of means. All others: Pearson chi square test.
K. I. KAPTEIN et al.
666Psychosomatic Medicine 68:662–668 (2006)
experience decreases in their symptoms without intervention.
Such a course of symptoms might be seen as a direct reaction
to the MI or even attributed to the MI due to symptom overlap
(e.g., fatigue) (40,41).
In conclusion, in our study, subjects with significant de-
pressive symptoms during hospitalization that increased dur-
ing the post-MI year had a significantly higher rate of new
cardiovascular events, which was not explained by any initial
measure of cardiac impairment. The other classes, although
having some differences in characteristics, showed no signif-
icantly impaired outcomes. Special attention should therefore
be given to those patients with increasing depressive symp-
toms during the post-MI year. Although in our sample this was
a relatively small subgroup, treatment of depression in these
cases may lead to both improvement in depression status and
to a reduction of new cardiovascular events.
The DepreMI was funded by a grant from the Netherlands Organi-
sation of Scientific Research (Zon MW, Grant 904-57-100).
The DepreMI study was conducted by the following persons and
Academisch Ziekenhuis Groningen: T. A. Spijkerman, R. H. S. van
den Brink, J. F. May, J. B. Winter, J. H. C. Jansen, H. J. G. M. Crijns,
Martini Ziekenhuis Groningen: J. H. Bennekers, F. van den Berg,
P. J. L. M. Bernink, R. B. van Dijk, M. G. Niemeyer, J. L. Postma,
L. E. J. M. Schrijvers, L. H. Takens.
Refaja Ziekenhuis Stadskanaal: K. de Vries, L. M. van Wijk.
St. Lucas Ziekenhuis Winschoten: T. R. Bouwmeester, A. van der
1. Denolett J, Sys SU, Brutsaert DL. Personality and mortality after myo-
cardial infarction. Psychosom Med 1995;5:582–91.
2. Frasure-Smith N, Lesperance F, Talajic M. Depression following myo-
cardial infarction: impact on 6-month survival. JAMA 1993;270:
3. Hance M, Carney RM, Freedland KE, Skala J. Depression in patients
with coronary heart disease: a 12-month follow-up. Gen Hosp Psychiatry
New Cardiac Events Compared to Class 1 (No Depressive Symptoms)
Cox Regression Analysis of Latent Classes 2 to 5 Risk of
HR 95% CIp
Class 2: mild depressive
Class 3: moderate depressive
Class 4: significantly
Class 5: significantly
Class 2: mild depressive
Class 3: moderate depressive
Class 4: sign. decreasing
Class 5: sign. increasing
Killip class ?2
History of MI
1.86 0.99–3.51 .055
1.42 0.84–2.40 .190
HR ? hazard ratio; LVEF ? left ventricular ejection fraction; MI ? myo-
Figure 3.Event-free survival for the five latent classes based on course of post-MI depressive symptoms.
COURSE OF DEPRESSIVE SYMPTOMS
667Psychosomatic Medicine 68:662–668 (2006)
4. Hemingway H, Marmot M. Evidence based cardiology: psychosocial Download full-text
factors in the aetiology and prognosis of coronary heart disease: system-
atic review of prospective cohort studies. BMJ 1999;318:1460–7.
5. Schleifer SJ, Macari-Hinson MM, Coyle DA, Slater WR, Kahn M, Gorlin
R, Zucker HD. The nature and course of depression following myocardial
infarction. Arch Intern Med 1989;149:1785–9.
6. Bush DE, Ziegelstein RC, Tayback M, Richter D, Stevens S, Zahalsky H,
Fauerbach JA. Even minimal symptoms of depression increase mortality
risk after acute myocardial infarction. Am J Cardiol 2001;88:337–41.
7. Lesperance F, Frasure-Smith N, Talajic M. Major depression before and
after myocardial infarction: its nature and consequences. Psychosom Med
8. Lesperance F, Frasure-Smith N, Talajic M, Bourassa MG. Five-year risk
of cardiac mortality in relation to initial severity and one-year changes in
depression symptoms after myocardial infarction. Circulation 2002;105:
9. van Melle JP, de Jonge P, Spijkerman TA, Gijssen JGP, Ormel J, van
Veldhuisen DJ, van den Brink RHS, van den Berg MP. Prognostic
association of depression following myocardial infarction with mortality
and cardiovascular events: a meta-analysis. Psychosom Med 2004;66:
10. Berkman LF, Blumenthal J, Burg M, Carney RM, Catellier D, Cowan
MJ, Czajkowski SM, DeBusk R, Hosking J, Jaffe A, Kaufmann PG,
Mitchell P, Norman J, Powell LH, Raczynski JM, Schneiderman N.
Effects of treating depression and low perceived social support on clinical
events after myocardial infarction: the Enhancing Recovery in Coronary
Heart Disease Patients (ENRICHD) Randomized Trial. JAMA 2003;289:
11. Spijker J, de Graaf R, van Bijl R, Beekman ATF, Ormel J, Nolen WA.
Duration of major depressive episodes in the general population: results
from the Netherlands Mental Health Survey and Incidence Study
(NEMESIS). Br J Psychiatry 2002;181:208–13.
12. Lewinsohn PM, Clarke GN, Seeley JR, Rohde P. Major depression in
community adolescents: age at onset, episode duration, and time to
recurrence. J Am Acad Child Adolesc Psychiatry 1994;33:809–18.
13. Solomon DA, Keller MB, Leon AC, Mueller TI, Shea MT, Warshaw M,
Maser JD, Coryell W, Endicott J. Recovery from major depression: a
10-year prospective follow-up across multiple episodes. Arch Gen Psy-
14. Mayou RA, Gill D, Thompson DR, Nicolas Hicks AD, Volmink J, Neil
A. Depression and anxiety as predictors of outcome after myocardial
infarction. Psychosom Med 2000;62:212–9.
15. Crowe JM, Runions J, Ebbesen LS, Oldridge NB, Streiner DL. Anxiety
and depression after acute myocardial infarction. Heart Lung 1996;25:
16. Goodman LA. The analysis of systems of qualitative variables when
some of the variables are unobservable, part 1: a modified latent structure
approach. Am J Sociol 1974;79:1179–259.
17. Goodman LA. Exploratory latent structure analysis using both identifi-
able and unidentifiable models. Biometrika 1974;61:215–31.
18. Lazarsfeld PF, Henry NW. Latent Structure Analysis. Boston: Houghton
19. Chen LS, Eaton WW, Gallo JJ, Nestadt G. Understanding the heteroge-
neity of depression though the triad of symptoms, course and risk factors:
a longitudinal, population-based study. J Affect Disord 2000;59:1–11.
20. Kendler KS, Eaves LJ, Walters EE, Neale MC, Heath AC, Kessler RC.
The identification and validation of distinct depressive syndromes in a
population-based sample of female twins. Arch Gen Psychiatry 1996;53:
21. Parker G, Wilhelm K, Mitchell P, Roy K, Hadzi-Pavlovix D. Subtyping
depression: testing algorithms and identification of a Tiered model.
J Nerv Ment Dis 1999;187:610–7.
22. Sullivan PH, Kessler RC, Kendler KS. Latent class analysis of lifetime
depressive symptoms in the national comorbidity survey. Am J Psychi-
23. Muthe ´n B, Shedden K. Finite mixture modeling with mixture outcomes
using the EM algorithm. Biometrics 1999;55:463–9.
24. Nagin D. Analyzing developmental trajectories: semiparametric, group-
based approach. Psychol Methods 1999;4:139–77.
25. Roeder K, Lynch KG, Nagin DS. Modeling uncertainty in latent class
membership: a case study in criminology. J Am Stat Assoc 1999;94:
26. Spijkerman TA, van den Brink RH, Jansen JH, Crijns HJ, Ormel J. Who
is at risk of post-MI depressive symptoms? J Psychosom Res 2005;58:
27. Spijkerman T, de Jonge P, van den Brink RHS, Jansen JHC, May JF,
Crijns HJGM, Ormel J. Depression following myocardial infarction:
first-ever versus ongoing and recurrent episodes. Gen Hosp Psychiatry
28. Beck AT, Mendelson M, Erbaugh J. An inventory for measuring depres-
sion. Arch Gen Psychiatry 1961;4:561–71.
29. Eysenck SBG, Eysenck HJ, Barrett P. A revised version of the Psychoti-
cism Scale. Pers Individ Dif 1985;6:21–9.
30. Wojciechowski FL, Strik JJ, Falger P, Lousberg R, Honig A. The rela-
tionship between depressive and vital exhaustion symptomatology post-
myocardial infarction. Acta Psychiatr Scand 2000;102:359–65.
31. Rubin DB. Multiple Imputation for Nonresponse in Surveys. New York:
John Wiley & Sons; 1987.
32. Rubin DB. Multiple imputation after 18? years. J Am Stat Assoc
33. Schaver JL. Analysis of Incomplete Multivariate Data. New York: Chap-
man & Hall; 1997.
34. Twisk J, de Vente W. Attrition in longitudinal studies: how to deal with
missing data. J Clin Epidemiol 2002;55:329–37.
35. Jones BL, Nagin DS, Roeder K. A SAS procedure based on mixture
models for estimating developmental trajectories. Sociol Methods Res
36. Kaplan EL, Meier P. Nonparametric estimation from incomplete obser-
vations. J Am Stat Assoc 1958;53:457–81.
37. Irvine J, Basinski A, Baker B, Jandciu S, Paquette M, Cairns J, Connolly
S, Roberts R, Gent M, Dorian P. Depression and risk of sudden cardiac
death after acute myocardial infarction: testing the confounding effects of
fatigue. Psychosom Med 1999;61:729–37.
38. Strik JJ, Lousberg R, Cheriex EC, Honig A. One year cumulative inci-
dence of depression following myocardial infarction and impact on
cardiac outcome. J Psychosom Res 2004;56:59–66.
39. Lauzon C, Beck CA, Huynh T, Dion D, Racine N, Carignan S, Diodati
JG, Charbonneau F, Dupuis R, Pilote L. Depression and prognosis
following hospital admission because of acute myocardial infarction.
40. Fielding R. Depression and acute myocardial infarction: a review and
reinterpretation. Soc Sci Med 1991;32:1017–28.
41. de Jonge P, Ormel J, Brink RH, van Melle JP, Spijkerman TA, Kuijper
A, van Veldhuisen DJ, van den Berg MP, Honig A, Crijns HJGM, Schene
AH. Symptom dimensions of depression following myocardial infarction
and their relationship with somatic health status and cardiovascular
prognosis. Am J Psychiatry 2006;163:138–44.
K. I. KAPTEIN et al.
668 Psychosomatic Medicine 68:662–668 (2006)